[Ccrtp-devel] is there a bug in sequential number wrap

[Yaniv Levy]

Hi All,
In the last few days I have been working on a crash bug that occurs consistently every time the input RTP packet stream sequential number wraps.  Although ccRTP code  handles sequential number wrap, I suspect it is not complete , although I havent found yet  the  bug.
I set up a simple test, using rtpsend and rtprcv. rtprcv prints sequential number of received packet ( this was not done in original code). Indeed, whenever seqnum wraped, rtprecv lost 2-3 packets. 
Is there a race between the access to the input packet linked list by rtpsession thread and "application" thread? i noticed read and write locks, but is this enough?
I am particularly concerned with the fact the read access to the linked list is not atomic. You have to call getTopSequentialNumber() first and then getData() with the return value. the linked list status may be changed between these two calls, particularly around wraps.

I'm still investigating, but has anyone encounter such problem?

Thanks,
Yaniv.